Draw Things Generate from Pipeline

This is the trigger at the end of the Draw Things wrapper's "advanced" flow. On their own, the pipeline nodes (Txt2Img Pipeline, Img2Img Pipeline, and the Add-Lora / Add-Control / Add-Custom modifiers) don't generate anything - they just build and edit a settings blob. DrawThingsGenerateFromPipeline is the node that takes that finished blob and actually fires the generation on your Draw Things app.

Think of it like a KSampler at the bottom of your graph: everything upstream describes what to render, and this node is the "go" button. The reason the pack splits it out this way is so you can construct a Draw Things config, then keep stacking modifiers onto it - a LoRA here, a ControlNet there, a custom field for some setting the other nodes don't expose - before anything runs. The basic Txt2Img/Img2Img nodes fuse "build config" and "generate" into one step; the pipeline route pulls them apart so the middle is editable.

How it works

Upstream nodes hand you a Python dict of Draw Things parameters - that's the "pipeline." This node takes that dict and POSTs it to the local Draw Things API (HTTP, port 7860). Draw Things runs the job on Apple Silicon (the whole point of this pack - it's faster than native ComfyUI on a Mac) and streams the image back as a ComfyUI IMAGE. It doesn't add or change any settings; whatever's in the pipeline when it reaches this node is exactly what gets rendered. So if a LoRA didn't apply, look upstream - this node just executes.

The inputs and outputs that matter

There's really only one input, and that's the point:

• pipeline (dict) - the settings blob coming out of a Txt2Img Pipeline / Img2Img Pipeline node, optionally after passing through Add-Lora, Add-Control, or Add-Custom nodes. This is the required and only input.

Output is generated_image (IMAGE) - the finished render, ready for a preview or save node.

That's deliberately minimal. All the actual knobs - model, prompt, steps, sampler, LoRAs, controls - live on the upstream nodes. This one just runs what it's given.

How to install it

It ships in the same pack as every other Draw Things node, so if you've installed one you have this. If not:

In Draw Things, open the Advanced tab and enable API Server (Protocol HTTP, Port 7860, IP localhost) - that's the endpoint this node calls.

Then the pack, via ComfyUI Manager (search ComfyUI-DrawThingsWrapper, install, restart) or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/JosephThomasParker/ComfyUI-DrawThingsWrapper
cd ComfyUI-DrawThingsWrapper
pip install -r requirements.txt

then restart ComfyUI.

Common issues & troubleshooting

"Where's my pipeline input coming from?" This node is useless by itself - it needs a pipeline dict from a Txt2Img Pipeline or Img2Img Pipeline node. If you just want a one-shot generation with no mid-graph editing, skip the pipeline route entirely and use the basic DrawThingsTxt2Img or DrawThingsImg2Img node instead; they do build-and-generate in a single step.

A LoRA/ControlNet you added didn't show up. This node doesn't touch settings - it only executes. Check that the Add-Lora / Add-Control node is actually wired into the pipeline chain feeding this node, and that the model/LoRA names match the Draw Things filenames exactly.

Re-running produces nothing new. ComfyUI only re-executes when an upstream ComfyUI input changes. Editing settings in the Draw Things app won't retrigger the graph - change something on a node (the seed works) to force it.

Connection fails. Draw Things has to be open with the API server enabled on port 7860, on the same machine. No app, no generation - this pack is a wrapper around a local app, not a cloud service.